Google's New Imagen 3 with Midjourney V6: Which is better?

At the Google IO 2024 event, Google announced a slew of brand-new products and huge AI updates. One of the major announcements was the brand new version of its text-to-image AI tool, Imagen 3.

Based on what they showcased during the announcement, there has been a significant improvement in visual quality. Imagen 3 has reached a level where it can easily compete with Midjourney v6.

But how do these two AI image generators compare side by side?

Let’s dive in and find out.

Prompt #1: Three women laughing

Three women stand together laughing, with one woman slightly out of focus in the foreground. The sun is setting behind the women, creating a lens flare and a warm glow that highlights their hair and creates a bokeh effect in the background. The photography style is candid and captures a genuine moment of connection and happiness between friends. The warm light of golden hour lends a nostalgic and intimate feel to the image

Both images look gorgeous, and the people in the frames are incredibly photorealistic. If I had to choose between the two, I’d still prefer the image generated by Midjourney. The specular reflection looks better, and the skin texture is smoother, giving a more natural feel to the candid moment.

Prompt #2: Bouquet of flowers

A large, colorful bouquet of flowers in an old blue glass vase on the table. In front is one beautiful peony flower surrounded by various other blossoms like roses, lilies, daisies, orchids, fruits, berries, green leaves. The background is dark gray. Oil painting in the style of the Dutch Golden Age.

Imagen 3 takes the win here. The softer and warmer tone of the overall image makes me want to hang it on my wall. While Midjourney also did a great job, it often uses wildly saturated colors that can take away from the naturalism of the result.

Generative AI Publication is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Subscribed

Prompt #3: Digital cartoon

A weathered, wooden mech robot covered in flowering vines stands peacefully in a field of tall wildflowers, with a small bluebird resting on its outstretched hand. Digital cartoon, with warm colors and soft lines. A large cliff with a waterfall looms behind.

Imagen 3 did a better job on this one. Despite trying several times, Midjourney continuously fails to adhere completely to the prompt — the robot does not stretch its hand and is not looking at the bird, which diminishes the emotional impact present in the first image.

Prompt #4: Human hands

A view of a person’s hand as they hold a little clay figurine of a bird in their hand and sculpt it with a modeling tool in their other hand. You can see the sculptor’s scarf. Their hands are covered in clay dust. a macro DSLR image highlighting the texture and craftsmanship.

I remember the days when everyone was talking about how bad AI image generators render hands and limbs. Today, almost all AI models have improved a lot in that aspect and the examples above represent that progress.

Comparing the two images, the sculptor’s hand is covered in clay dust in the Midjourney-generated image, while it’s very clean in the Imagen 3 version.

Prompt #5: Text rendering on a speech bubble

A single comic book panel of a boy and his father on a grassy hill, staring at the sunset. A speech bubble points from the boy’s mouth and says: ‘The sun will rise again’. Muted, late 1990s coloring style

In this example, to be fair to Midjourney, I tried generating the image five times but failed to get the correct text rendered. Even after adding quotes to the text to fit Midjourney’s text rendering rules, it wasn’t able to render the text properly.

Prompt #6: Fine details

Elephant amigurumi walking in savanna, a professional photograph, blurry background

Both results are stunning, with mind-blowing levels of detail on the yarn loops. It’s easy to mistake them for real photographs. However, if I had to choose which one is better, I’d say the one from Midjourney edges out Imagen 3 in this case. Do you agree?

Prompt #7: Text rendering with feathers

Word “light” made from various colorful feathers, black background

This is a good example of just how better Imagen 3 is with text rendering capability. It was a nice try from Midjourney but the result isn’t that legible and contains unwanted artifacts.

This is a cherry-picked result from Imagen, though. I don’t know how many times they had to generate the image with the same prompt to get that awesome image.

Prompt #8: Illustration

Detailed illustration of majestic lion roaring proudly in a dream-like jungle, purple white line art background, clipart on light violet paper texture

Comparing the two images, Imagen 3 demonstrates more consistency as a line art piece, with colors much closer to the requested light violet compared to Midjourney’s result. Both look very cool, though, and it’s impressive to see AI handle various art styles.

Prompt #9: Claymation scene

Claymation scene. A medium wide shot of an elderly woman. She is wearing flowing clothing. She is standing in a lush garden watering the plants with an orange watering can

Both images adhere to the prompt, but the one from Imagen 3 looks more polished. In the Midjourney version, the elderly woman’s hand holding the watering can doesn’t look quite right, and the water doesn’t come out directly from the can’s spout.

Prompt #10: Creatures

Photographic portrait of a real life dragon resting peacefully in a zoo, curled up next to its pet sheep. Cinematic movie still, high quality DSLR photo.

While Imagen 3 has improved significantly in generating creatures, Midjourney is still the king in this category. Just look at how cute the dragon and sheep look together in the Midjourney image.

Share

How to get access

Head over to Google’s official blog post of Imagen 3 and click on the “Sign up to try on ImageFX” button.

ImageFX is part of Google’s test kitchen for its AI tools.

You can also request access to Imagen 3 from the ImageFX dashboard.

Okay, that’s about it. I hope you found this comparison article helpful. If you want me to do a comparison of Imagen 3 against other image generators like OpenAI’s Dall-E 3 or Adobe Firefly 2.0, let me know.

Final Thoughts

Overall, it’s great to see these two image models performing really well. The images are very detailed, coherent, and overall stunning.

From an aesthetic perspective, I still find Midjourney superior, but we have now reached a saturation point in text-to-image models the text rendering is on the OpeAI’s Dall-E 3 level.

While it’s important to keep in mind that the example images are cherry-picked by Google and may not be fully representative of Imagen 3’s performance once it’s publicly available, I must admit that I’m impressed by what I’ve seen so far.

‍